|
| ||||||||||||||||
| ShaoLin Aptus 2.0 Installation and Operations Guide | ||
|---|---|---|
| Prev | Chapter 12. Tips on the Linux environment | Next |
There are quite lot of file systems support in Linux, the most common type is ext2, others such as ext3, xfs, jfs, and reiserfs are getting more common and they will likely enter the standard Linux kernel. This chapter is to address some experience of the author using those file systems with Aptus and Linux administration. It also cover some performance and reliability aspects on using those file systems.
People are moving to Ext3 because of the journaling feature. Journaling enable fast file system recovery on unclean shutdowns. Ext2 and Ext3 are backward compatible, you can move your Ext2 by using the tune2fs utility which comes with most of the decent Linux distributions. To convert a file system from Ext2 to Ext3, unmount your file system and do the following.
The -j option will create journal block in your existing Ext2 file system which in turn a Ext3 file system. To find out more details about journal block size, please checkout the man 8 tune2fs. Ext3 has only medium performance, however, the reason to use Ext3 is because of compatibility. It also have extensive tools for rescue and debugging. It is save to use Ext3 to trade performance with protection.JFS is called journaling file system and is developed by IBM, it has journaling feature and reasonable performance. It is a file system comes from a mature operating system IBM AIX . Currently, JFS is not in the standard Linux kernel, you may find difficulties in getting help or rescue JFS in Linux. The Mandrake Linux distribution ships JFS in their stock 2.4.18 Linux kernel, so if you want to use JFS, you have to use Mandrake Linux.
XFS is developed by Sgi, it is a high performance journaling file system original from the IRIX operating system. Currently, Mandrake Linux ship XFS with its stock kernel. This file system have higher average performance than Linux Ext3 and Ext2 with journaling feature. XFS have some special features like guarantee I/O rate and fast lookup on large directories, it also support very large file size and is on-disk compatible to IRIX version XFS. It also support a block size of 64K which benefit on large file systems.
A pure Linux based high performance journaling file system. At the date of publication, reiserfs is in version 4 . Our experience shows reiserfs version 4 has the highest performance out of the above file systems, it is useful on storing lots of files in a directory without experience delayed file lookups. The problem of reiserfs is its block allocation make it easy to corrupt on unclean shutdown . If you really concern about reliability (such as storing data files or user home directories), please use other file systems. But I still recommend using reiserfs on /usr and other non frequently changed file systems which they have less likely to be unsynced.
LVM is an enterprise level volume management driver. It is useful on dynamically resize partitions without have to backup or alter partition information. To use LVM, you must create the LVM volume group and logical volume at setup time. It is suggested to use LVM on home directory servers where storage capacity has a high demand. LVM itself is not a file system, in fact it is a multi-disk block device driver. Other than scalability of capacity, it also provides snapshot features of file systems which is useful in making incremental backup of system. LVM can also strip your data into multiple disks where performance can be increased over data stripping. See more details on LVM How-to.